Abstract
Overtourism can place pressure on local communities, infrastructure and the environment, making the assessment of tourism pressure an important challenge for sustainable destination management. This study expands on overtourism risk assessment across Italian provinces by conducting a systematic comparative analysis between Multi-Criterion Decision-Making (MCDM) sorting methods, AHPSort II and FlowSort, and unsupervised clustering techniques, specifically K-Means and Ward hierarchical clustering. Building on previously reported MCDM classifications, this study adds comparisons across all four methods, pairwise agreement statistics, internal cluster validation and indicator-dependence diagnostics. While MCDM methods enable preference-based categorization against predefined risk profiles, clustering techniques offer a similarity-based benchmarking perspective. The comparison highlights how decision preferences, data representation and assignment mechanisms contribute to differences between classifications. The empirical analysis reveals 91.59% agreement between the two clustering approaches, corresponding to 98 out of 107 provinces. Combining outranking, preference interpolation, and distance-based clustering highlights the complementary nature of these tools: clustering identifies groups with similar indicator profiles, while MCDM delivers structured, preference-based risk categorization. Agreement between methods does not establish external validity, as no independent reference classification of overtourism risk is available. Ultimately, contrasting preference-based sorting with similarity-based clustering enhances decision framework transparency, helping policymakers to identify provinces requiring closer investigation and informing sustainable tourism planning.